networkdynamics
adapters.networkdynamics
NetworkDynamics.jl backend adapter for SimulationExperiment.
Uses pyjulia (tvbo.adapters.julia) to execute generated Julia code and return full Julia objects alongside a TVBO TimeSeries.
Attributes
| Name | Description |
|---|---|
| REQUIRED_PACKAGES |
Classes
| Name | Description |
|---|---|
| NetworkDynamicsAdapter | Adapter for running SimulationExperiment via NetworkDynamics.jl (pyjulia). |
NetworkDynamicsAdapter
adapters.networkdynamics.NetworkDynamicsAdapter(experiment)Adapter for running SimulationExperiment via NetworkDynamics.jl (pyjulia).
Inherits metadata processing from BaseAdapter. The prepare_context() method pre-computes all template variables so Mako templates stay clean.
Methods
| Name | Description |
|---|---|
| build_node_positions | Build (n_t, n_nodes, n_cv) position array from simulation data. |
| get_fixed_nodes | Return set of node IDs with static dynamics (no state variables). |
| get_initial_positions | Extract initial (x, y, …) positions for ALL nodes from YAML. |
| get_node_metadata | Extract per-node metadata: dynamics name, parameters, type. |
| refuse_unrenderable | Raise where the emitted Julia would quietly integrate something other than what was declared. |
| run | Run simulation using NetworkDynamics.jl. |
build_node_positions
adapters.networkdynamics.NetworkDynamicsAdapter.build_node_positions(ts, ctx)Build (n_t, n_nodes, n_cv) position array from simulation data.
For free nodes: positions come from the coupling-variable columns of the properly shaped (time, variable, node) DataArray. For fixed nodes: positions are constant (from YAML parameters).
get_fixed_nodes
adapters.networkdynamics.NetworkDynamicsAdapter.get_fixed_nodes()Return set of node IDs with static dynamics (no state variables).
get_initial_positions
adapters.networkdynamics.NetworkDynamicsAdapter.get_initial_positions()Extract initial (x, y, …) positions for ALL nodes from YAML.
For free (dynamic) nodes: positions come from per-node state overrides (legacy initial_state arrays are also supported), at the indices marked coupling_variable=True. For static (fixed) nodes: positions come from node parameter values (in parameter-definition order).
Returns shape (n_nodes, n_coupling_vars).
get_node_metadata
adapters.networkdynamics.NetworkDynamicsAdapter.get_node_metadata()Extract per-node metadata: dynamics name, parameters, type.
Returns {node_id: {'dynamics': str, 'params': dict, 'static': bool}}.
refuse_unrenderable
adapters.networkdynamics.NetworkDynamicsAdapter.refuse_unrenderable()Raise where the emitted Julia would quietly integrate something other than what was declared.
The edge and experiment templates carry no delay path, so a delayed coupling is not lowered — it is dropped, and the run returns a well-formed trajectory of the undelayed network. They carry no observation path either, so a declared monitor is dropped the same way and the result arrives with an empty observations. Refusing is the same contract the Brian2 adapter states for the forms it cannot lower: a clear error beats a plausible answer to a different question.
run
adapters.networkdynamics.NetworkDynamicsAdapter.run(**kwargs)Run simulation using NetworkDynamics.jl.
Returns:
ExperimentResult Simulation results with named dimensions and coordinates. Extra attributes: sol, graph, edge_data, vertex_data.